Potential biases due to geocoding error in spatial analyses of official data.

Geospatial methods have been used extensively to examine associations between alcohol outlet density and various harms; however, the literature offers too little methodological detail to assess possible geocoding biases in these studies. We used New Zealand liquor licensing and crime data to assess geocoding error. For the year with the best data, 69% of offences could be accurately mapped (91% of those in urban areas, 38% in rural areas). There was considerable urban-rural variation in the accuracy and specificity of location data. If similar error exists in other jurisdictions, previous findings may be biased. Greater consideration should be given to the effects of data quality in geospatial studies, and geocoding methods should be reported explicitly.